Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

MSWIn, 9i -> 11G Upgrade, 100% CPU Usage, No Workload

159 views
Skip to first unread message

1001...@gmail.com

unread,
Apr 16, 2013, 12:08:03 PM4/16/13
to
Hi All,
I have been running fully patched 11G instances in my development and staging environments, MSWin2003, for over 2 years, never an issue.

Finally, we do the production upgrade with a MSWin2008 on HyperV and within 30 minutes CPU is at 100% with only 2 connections made and no workload, nothing current in v$sql*. Rollback......

Repurpose my hot standby (MSWin2003, bare metal), temporarily alias it as the production database and the same problem occurs. After a physical reboot of the server, issue re-occurs within 10 minutes.

Oracle Support says turn off all jogs, particularly the DEFAULT_MANAGEMENT_PLAN (?). Repeat the test problem occurs. I believed I had identified the cause, possible bad statistics in a schema.

Move the machine to a test environment with the identified client. Test the client with a connection and simple report to my staging database (STG), which immediately consumes 100% CPU. I test the client against my "prod now test" databases, with an exact copy of what had been done to cause the problem in production. Manually, create the db, datafiles and users, import from 9I. Save and calculate stats at various different times. However, I cannot consistently reproduce the problem. I am successful on 2 out 10 tries.

All occurrences or the problem on new databases will repeat almost immediately after a bounce, STG will stay up for a day or two or more before it occurs again.

STG has had lots of drop and create users and imports but the database itself is over 1 year old.

We do not run OEM or emctl, I'm a command line kind of guy.

I'm baffled, any ideas? I have opened an SR but .....

TIA
Dave

joel garry

unread,
Apr 16, 2013, 8:31:23 PM4/16/13
to
Sounds like you should run a 15 minute statspack at good and bad times
and see what that has to say about it. Google Jonathan Lewis
Statspack if you need more info. Something is going plan-crazy I
would guess.

jg
--
@home.com is bogus.
http://finance.yahoo.com/news/ceos-earn-354-times-more-than-average-worker-203748434.html

1001...@gmail.com

unread,
Apr 17, 2013, 11:25:49 AM4/17/13
to
Hi Joel,

> Sounds like you should run a 15 minute statspack at good and bad times
> and see what that has to say about it. Google Jonathan Lewis
> Statspack if you need more info. Something is going plan-crazy I
> would guess.

At one point I thought bad stats too but....
We've taken AWR snapshots and reports multiple times. All the sql runs normally.
We have imported/exported stats during normal operation and when the above happens. Applying the good and bad stats seems to have no bearing on when the problem happens. Notice I say seems :(

Thanks for your thoughts
Dave

joel garry

unread,
Apr 17, 2013, 11:37:50 AM4/17/13
to
Just to be clear, when you say "machines" you don't mean more HyperV?

jg
--
@home.com is bogus.
http://searchoracle.techtarget.com/news/2240181880/Oracle-competition-is-worth-a-look-Collaborate-attendees-say

1001...@gmail.com

unread,
Apr 19, 2013, 3:41:08 PM4/19/13
to
Hi Joel,
the MSWIN2008 machines are Hyper-V on dedicated metal, one VM to one physical machine. We have run the MWIN2008 on bare metal but it made no difference

The MSWIN2003 machines are bare metal.

All configurations experience the problems,.... sometimes :(

Dave

Mark D Powell

unread,
Apr 20, 2013, 1:56:18 PM4/20/13
to
Windows 2008 or Windows 2008 R2? What specific SP? I see no point in using Hyper-V if you are only going to run one environment under it. Why add the extra overhead of a VM?

What is the total memory in the machine? How much was allocated to the OS? How big a page file was allocated?

I would pull the Oracle Windows installation manual and verify any listed required OS settings were in fact made.

HTH -- Mark D Powell --

John Hurley

unread,
Apr 20, 2013, 7:55:09 PM4/20/13
to
So what processes ( oracle or not oracle ) are using cpu?

If not oracle related processes then what are they?

If oracle related processes ... what processes? If not a background
kind of process try running a 10046 trace on them ... put trace thru a
resource profiler.

Look at tanel poders set of diagnostic scripts ( snapper etc ) ... are
you on cpu or waiting ... etc.
0 new messages